FanPost

Penguins Defy Predictive Statistics, alternatively Corsi Schmorsi

Our Pittsburgh Penguins are struggling to out shoot components, seeing Corsi and Fenwick percentages drop like a rock, yet they keep winning. The concern is that Corsi and Fenwick are correlated with winning so odds will eventually catch up, especially in the playoffs.

Statistics can certainly show why teams won or lost: who possessed and shot the puck more (Corsi/Fenwick) and who had better goal tending, quality shots, and luck (PDO). But they can also be deceiving, such as when the Pens' power play tried to keep the puck without shooting a couple years ago. Yes, longer amounts of time in the offensive correlate with scoring, but it's an effect of strong possession and chances rather than the cause. Statistics are complicated, as is the game, but at the end of the day hockey comes down to scoring more than the other guy. The team that shoots more with better save and scoring percentages is going to win.

This is interesting, so let's dive right in. Fenwick is important as a proxy for possession and PDO is important as a proxy for quality (scoring and save) and luck, although we don't know which. Below we see a steady decline in 5x5 FF% (possession when the game is close).

Pittsburghpenguins2013-2014cumulative5v5closeff_zps3aaff8eb_medium

A team with 65% FF is not sustainable (there is some luck in there too), but the trend from 55% to 50% indicates the team is getting worse. If you believe in the ol' eye ball test, you should agree with this. Letang, Martin, and Dupuis have been out of the lineup, not to mention Neal, Malkin and Bennett. This matches up with the PDO chart. Fortunately FF was so good at the beginning of the year, because PDO was awful. To put this in perspective, every 1 point over 100 in PDO is roughly a 10% chance of outscoring the other team (they shoot and allow 57 shots per game on average, league combined goals per game are 5.3 this year = 57*(1.01-1)/5.3). Throwing a large number of pucks at the net hides quality scoring chances from just trying to get a good bounce, so PDO generally equals 100 over time.

Pittsburghpenguins2013-2014rolling10-game5v5pdo_zps95b738cb_medium

We're right around 100, meaning we're doing ok. But, PDO and FF combine to tell the story. If we're at 100 and the other team keeps shooting more and out possessing, then odds are the other team will win. Going back to the cumulative FF chart, we know a decreasing one means we're seeing negative numbers (bringing down the average). Below is a chart showing the rolling average and what happened when the three highest impact injuries occurred this year.

Chart_zpsf9a0f678_medium

Don't get carried away with the spike after Martin is injured. We know he's a quality defenseman. Check out what happens when we lose the top two puck moving defensemen and Crosby's winger. That seems to be the turning point related to possession.

The logical reply, in the words of the great Natasha66, becomes, "Corsi Schmorsi." That turns out to be spot on, at least for the Pens. Digging into it more absolutely shocked me. Running a regression against points per game, we have the following:

Coefficients

Standard Error

t-stat

P-value

Intercept

-770.737

90.62765

-8.50443

4.06E-09

FF%

1.787901

0.345056

5.18148

1.87E-05

PDO

0.756044

0.090171

8.384533

5.38E-09

R squared came out to 79%, which is the percentage of variance explained by FF and PDO. The standard error, t-stat, and p-value show the results as statistically significant. We can apply these coefficients to see what number of points the Pens should have. Here it is for all 30 teams sorted by expected points:

Team

Games Played

Actual Points

Expected Points

Difference %

Boston

66

93

98

-5%

St. Louis

66

97

95

2%

Anaheim

66

93

91

2%

Chicago

66

90

91

-1%

Colorado

66

91

85

7%

Los Angeles

67

82

85

-3%

San Jose

67

93

84

11%

Columbus

66

74

83

-11%

Tampa Bay

66

77

82

-6%

Minnesota

66

79

78

1%

Dallas

65

74

78

-5%

Phoenix

67

73

77

-5%

Philadelphia

65

73

75

-3%

Pittsburgh

65

92

75

23%

Toronto

68

80

75

7%

Ottawa

65

68

74

-9%

Detroit

65

71

74

-4%

Vancouver

68

70

73

-4%

NY Rangers

67

74

71

4%

New Jersey

66

71

70

2%

Washington

67

70

70

1%

Carolina

66

67

69

-3%

Winnipeg

67

69

69

0%

Montreal

67

77

68

14%

Florida

66

59

64

-8%

Edmonton

67

54

60

-10%

NY Islanders

67

59

60

-2%

Nashville

66

66

58

14%

Calgary

66

59

57

4%

Buffalo

66

46

53

-13%

The bottom of the list makes sense as does the top. Looking at the differences between actual and expected points, we see the teams outperforming their predicted results (Pittsburgh, Montreal, Nashville) and underperforming (Buffalo, Columbus, Edmonton). Note that Pittsburgh is by far the largest outlier.

Some of the reasons that could account for this are their injury situation, league leading special teams, and winning shootouts (5-2, which accounts for 1.5 points). However, these effects should be captured in FF and PDO. I hate to be anti-climatic, but I really don't know how the Pens keep winning with such an awful FF. Maybe it's schedule or luck. Maybe Bylsma knows something everyone else doesn't.

What needs to happen is something we all knew to begin with. They need to get healthy. Once they get healthy, they need to learn to play together to get that FF and PDO number up for the playoffs. Otherwise, with the current team and injury situation, a deep playoff run is not going to happen.

*Data is from Extraskater.com, Stats.HockeyAnalysis.com, and NHL.com as of 3/12/14

The content expressed in fanposts does not necessarily reflect the opinions of the staff here at Pensburgh.com. FanPosts are opinions expressed by fans of various teams throughout the league but may be more Pittsburgh-centric for obvious reasons.